Theoretical accounts of LI

From a theoretical perspective, LI is either considered as a deficit of a cognitive mechanism

specialised in production of linguistic representations (domain-specific approach) or as the result of a more generalised cognitive weakness (domain-general account). The domain specific approaches are grounded on the generative theory (Chomsky, 1965) and have, therefore, employed concepts from it.

The main idea here is quite specific: the “innate” grammatical system is flawed, leading to LI.

However, more than one “representational” sub-theories have emerged, which differ on whether their interpretation “covers” a specific grammar sub-system or provides an explanation involving a more complete range of morphological impairments.

Within these different approaches exist further theoretical sub-distinctions (Stavrakaki S., in Pontikas G.; 2016). For example, even within representational accounts there is no consensus as to what the fundamental nature of the observed morphological difficulties across languages is. For instance, Clahsen (1991) argues for an agreement deficit based on German data –further explained in the next chapter-whereas Rice & Wexler (1996) maintain that there is a tense deficit based on English data.

Based on Rice’s theory, which is termed Extended Optional Infinitive Account, the lack of obligatory tense marking (finiteness), is the characteristic of an -early- developmental stage of typical language development, which in SLI, is ‘extended’.

As regards the agreement deficit approach (Clahsen, 1991), it is applied in Germanic languages but may not account for LI in non-null-subject languages, where agreement deficits are much rarer. An additional, more recent, “representational” approach has been proposed by Van der Lely (2000), which aimed to account for the errors of the previously mentioned “Grammatical SLI” children, which is a category of children with specific, syntactic errors. Based on her theory a deficit of the Computational Grammatical Complexity, i.e., the computational system responsible for grammatical operations, is responsible for problems with structure-dependent relationships, observed in marking tense and

23 agreement, case assignment, as well as “movement operations”. Furthermore, this approach has been further applied to explain other frequently observed morphological behaviours of LI children, such as omission of plural or third person -s. Based on the Computational Grammatical Complexity theory, errors such as omissions are the result of an inability to verify the grammaticality of a given word within a sentence, in the presence of adequate memory skills (Van der Lely & Christian, 2000, Gopnik, 1990). Further examples include the omission of plural-s within compound words, as well as the ability to interpret “movement operations” during wh-questions, whereby the question word needs to be placed at the beginning of the word. With regards to such tasks, Marinis & Van der Lely (2007) found that G-SLI children did not rely on syntactic information but rather use semantic data to analyse and interpret those types of questions.

Finally, an approach that may be classified as a ‘representational’ account, has been proposed by Ullman (2004). The so-called Declarative-Procedural model emphasises the application of

grammatical knowledge through a dual system: declarative knowledge, which governs everything that is learned in an automated, subconscious way, such as lexical information, and procedural knowledge, which is involved in conscious and active learning, such as the rule-governed, implicit, grammatical morphology. According to this theory, LI-children’s procedural memory is deficient, whereas declarative memory remains relatively intact. This explains the fact that learning the rules of

morphology, such as in agreement inflection, is highly problematic, while at the same time, it accounts for the existence of a few instances of compensated, ‘correct’ grammatical structures that are based on declarative memory.

Research evidence supporting that procedural, implicit learning is, indeed, problematic in children with LI stems from studies, whereby LI-children presented with slower reaction times compared to controls during a serial reaction time task (Tomblin, Mainela-Arnold & Zhang, 2007), as well as research using classical artificial grammar learning tasks showing that LI children were less able to implicitly learn the rules of a novel grammar structure based on paradigms from that grammar (Alt, Plante & Creseure, 2006; Evans, Saffran, & Robe-Torres, 2009). This ability has been further tested through alternative measures, such as statistical testing, where individuals are required to track patterns during transmission of different stimuli, such as syllables and tones, in order to extract regularities; the fundamental assumption of these studies is that, typically, infants and adults are easily able to detect familiar vs novel patterns of presented stimuli. This ability has been linked to being able to identify the word boundaries when listening to streams of sounds and, thus, learn new words.

Evans, Saffran, & Robe-Torres (2009), set out to explore how LI children perform on such tasks of implicit learning, and whether their performance depended on exposure and type of stimulus

(linguistic or not). Their conclusions were that LI’s children “computational” ability to track statistical regularities in order to identify word boundaries is less developed than in typically developing

children, that it depends on the time of exposure to the stimuli, and that this difficulty does not only

24 involve linguistic (i.e., speech) but, also, non-linguistic stimuli, such as tones. Overall, they conclude that the observed deficits may be explained by an adapted version of the representational account presented by Ullman that would include other aspects of implicit learning, such as word learning.

On the other hand, processing accounts vary from those proposing general processing capacity deficits to be the main cause of SLI (Leonard, 2000) to the ones suggesting that limitations of specific

subsystems, such as phonological memory, are responsible for the observed deficiencies (Gathercole

& Baddeley, 1990).

The Morphological Richness account and the Surface Account are two closely related processing theories both postulated by Leonard (2000). The main idea of both accounts is that LI is caused by difficulty to perceive specific morphological characteristics that are more transparent in certain languages compared to others, as indicated by many cross-linguistic studies. More specifically, the Surface Account links LI deficits with a shortage of the auditory processing system to perceive less salient information, especially in languages with less clear contrasts between strong and weak surface characteristics, such as syllabicity, duration of syllables/ morphemes etc. This leads to weaker

morphological representations, especially in languages with less salient grammatical characteristics e.g., English compared to Spanish. In the same vein, based on the Morphological Richness Account, languages with richer inflectional systems provide opportunities for more consistent input to different morphological structures, thus, facilitating the creation of correct grammatical representations.

With regards to the domain-general (processing) accounts, they have been supported by studies evidencing that LI children demonstrated slower response times during linguistic and non-linguistic tasks (Gillam, Gillam, Fargo, Olszewski, Segura et al., 2017; Leonard, Weismer, Miller, Francis, Tomblin et al, 2007). Within this context the required processing time is synonymous to processing capacity; also, each presented task might involve several stages that contribute collectively to the overall required measurable response time.

To illustrate how processing accounts interpret disordered patterns compared to the representational theories from a practical perspective, we will refer to the previous example of s-omission. Based on this approach, children omit plural or third person -s because they do not perceive this sound-morpheme and, therefore, cannot acquire the corresponding grammar rule. In addition, they have difficulties expressing it due to phonological constraints at the short-term memory and processing level (Van der Lely, 2000; Leonard, 2000).

25 1.3 Language Impairment in German monolinguals

Given that this study entails children with German as an additional language it is important to highlight a few linguistic characteristics of German, as well as areas that are a frequent cause of concern in LI children.

The German language makes extensive use of grammatical cases and inflections -with different possible variations- and has a variable but rule-governed word order. For instance, where nouns are concerned, there are three possible ways to mark the gender (through articles) of nouns, four different cases and four allomorphs to mark plural, whereas verbs are inflected not only for person but, also, for number and tense. Furthermore, verb is always placed as the second element in a sentence, unless a compound verb or subordinate clause is used, in which case the verb goes at the end of the sentence (Leonard, 2000). In addition, due to its relevance to our study specific reference should be made to a general “sentence bracket” system, which is characteristic of German syntax, whereby the predicate in the main clause has infinite (non-conjugate) parts in addition to the finite verb (the conjugated verb form). The left sentence bracket typically contains finite verbs or subordinating conjunctions. Mastery of this sentence bracket/ assembly system is considered especially hard for LI children (Shulz &

Tracy, 2011).

1.3.1 Grammatical disorders in German

The morphological characteristics described in the previous section pose significant strains on young, language impaired children who struggle to master verb morphology, verb order and agreement, case markings and so on. There is data showing that agreement marking is one of the main clinical markers of LI in German (Chilla, Rothweiler, & Clahsen, 2012) confirming one of the earliest accounts of LI in German, i.e., the agreement deficit hypothesis, based on which the ability to match features of

different syntactic features within a sentence, such as verb and object, is impaired in LI (Clahsen, 1991). Alternative accounts include the ‘Extended Optional Infinitive’ hypothesis (Rice, Wexler &

Cleave, 1997), as well as the “deficient Complementizer Phrase” (Hamann & Plunkett, 1998) hypothesis. In support of the former theory, there was evidence that LI children omitted copula and used non-finite verbs (e.g., infinitives) in finite contexts more frequently than typically developing children. Hamann (1998) used the CP hypothesis to explain phenomena, such as verb-final placement,

“target-inconsistent” subordinate clauses and more frequent use of finite verbs, which she observed by reviewing the spontaneous speech data from numerous children with LI. Finally, there is evidence that LI in German manifests itself, also, as a decreased ability to understand several question forms, as well as meanings of verbs (Schulz & Tracy, 2011).

26 1.3.2 Phonological and phonetic disorders in German

Speech sound disorders may or may not accompany children with LI. Of great value in this domain is the work of Fox-Boyer and her colleagues (2016), who identified and classified the speech patterns of 276 children with suspected SSD or LI following Dodd’s classification system (1995), i.e., looking at whether speech output was typical but delayed or uncommon (“disordered”) for the children’s age, and/ or whether speech sound production was inconsistent across different trials. Based on her findings, most of the children presented with a delayed rather than a disordered phonological pattern.

Most delays concerned the correct realisation of sounds /ʃ/ and /ç/, velar stops and consonant clusters, whereas most disordered patterns included contact assimilation (/tʁ, dʁ/ → [kʁ, gʁ]), and stopping or general difficulties with fricatives and affricates. Finally, in addition to the disordered phonological patterns, the existence of an inconsistency rating higher than 40%, as suggested by Dodd, along with overall compromised intelligibility constitute indications of phonological disorder in German (Neumann, Rietz & Stenneken, 2016; Albrecht, 2017)

1.3.3 Semantic-lexical disorders in German

LI in German is often associated with lexical deficits from a very young age. It often manifests itself through the late onset of vocabulary production, the existence of a small lexical inventory or the slow rate of acquisition of new words (Siegmüller & Kauschke, 2006). In many cases, these difficulties continue in early childhood as receptive vocabulary deficits or word-finding difficulties. For instance, LI preschoolers were found to perform worse at a fast-mapping task compared to their TD peers. More specifically, they found it harder to define verbs and recalled less semantic features of the presented items (Skerra, 2009). These difficulties may often lead to the use of generic words, repetitions and circumlocutions. Furthermore, there is an uneven acquisition of nouns compared to verbs and functional words (Rupp, 2013).

In addition, as children become older these semantic deficits frequently influence the understanding and the acquisition of new concepts or more complex sentences. For instance, a study employing event-related brain potentials concluded that LI children exhibited weaker lexical–semantic representations of verbs and their selection restrictions during an auditory comprehension task compared to their TD peers (Sabisch & al, 2006). Conclusively, language-impaired, monolingual German-speaking children are often found to present with semantic deficits that may also coexist with other morphological or phonological difficulties.

27 1.4 Language Impairment in bilinguals

1.4.1 Alternative terminology

Language impairment in bilinguals is often, also, referred to in literature as Primary Language Impairment (PLI) or linguistic disorder and it is differentiated from language difference, which is the

“"the result of the normal process of second language acquisition, and its impact on the development of the second language" (Gillespie, 2015) and describes a typical pathway to bilingual linguistic mastery. In this study, the terms LI, PLI and linguistic disorder will be used interchangeably. While the definition of LI in monolinguals is generally standard, albeit occasional changes, such as the recent emergence of the term DLD (Developmental Language Disorder) reference (Bishop, 2017), consensus as regards the criteria of LI in bilinguals is far less clear. For instance, although some researchers argue that criteria should include performance below one standard deviation on a standardised linguistic assessment on the child’s L2, such as in English (Cleave, Girolametto, Chen & Johnson, 2010), other question the validity of using such normative data and instead propose the use of parental reports (Tsybina & Eriks-Brophy, 2010). The issue of the assessment of LI in bilinguals will be discussed further in chapter 1.8.

A summary of the main criteria of PLI in bilinguals could include: - non-verbal abilities within the average range; absence of oral motor, neurological or developmental difficulties, such as autism; -uneventful medical history and intact hearing and – parental concern regarding development of L1 (and L2); (Hemsley, 2015).

1.4.2 Linking theoretical accounts of LI to bilingual acquisition

As has been discussed earlier, a lot of researchers have pointed out that children with LI and (s equential) bilinguals tend to display commonalities as regards the observed errors in their expressive language. Such commonalities include common developmental patters in the types of error types the make, for instance, in expressive grammar. This makes it quite hard to distinguish between diagnostic features typical of “bilingual” LI and typical L2 acquisition. Furthermore, it is generally accepted that direct comparisons between these groups may result in “diagnostic confounds” and are associated with a risk if they are used as clinical markers; Also, the great variety of the followed methodologies, as well as the diversity of linguistic characteristics of the participants which support a common developmental pathway between LI and bilingual acquisition, further confirm this “risk”.

Taking into consideration the main LI accounts presented earlier, and extending them to bilingual acquisition, (Orgassa, 2009) mentions the following two alternatives: a) if one considers LI to be caused by deficits of their linguistic systems (see representational deficit) it may be expected that

28 bilingual LI children would present with similar profiles with their LI peers in each of their first languages (Paradis, 2007); this prediction has been based on the Disruption-within-Delay account (Rice 2003), which posits that there are some specific morphological elements that are significantly more affected in LI and are, thus, much harder to master both by mono- and bilingual LI children. This unique morphosyntactic deficit, which is in discrepancy with their overall linguistic development, may differentiate children with LI from their younger typically developing peers.

b) One the other hand, if LI is to be attributed to a general processing deficit, one may expect bilingual LI children to be delayed both compared to their typically developing bilingual peers, as well as their monolingual LI peers in each of their languages. This estimate stems from the Generalised Slowing Hypothesis (Kail, 1984) based on which a more global language delay is anticipated as a result of limited processing speed of the linguistic input required for grammar generation. Within this theoretical framework error types are to be of the same type between LI and typically developing children, but LI-bilinguals may present with more severe delays when compared to unimpaired-bilingual and monolingual LI children. Due to the already described effect of specific perceptual characteristics of different languages on certain grammar structures, this account has been extended to incorporate the importance of being able to perceptually finetune to the least salient features of the input languages (Surface Account; Leonard, 2000).

1.4.3 A brief description of the data regarding the profiles of bilinguals with LI per linguistic area

1.4.3.1 Speech sound disorders in bilingual children

Due to the already mentioned difficulties in the study of disordered speech of bilinguals, such as the lack of reference norms and the variability of sound systems there are few studies on this domain.

There are indications of both commonalities and differences as regards the types of errors made by bilingual children compared to monolinguals. For instance, based on some studies the speech output between different groups of bilingual children with SSD did not differ significantly as regards the place or manner of articulation or consonant accuracy (Goldstein, 2015). On the contrary, other data suggested that bilinguals displayed a higher occurrence of certain types of errors compared to monolinguals. Also, different types of errors have been found to be distributed across different

languages (Goldstein, 2015). As mentioned earlier, phenomena such as cross-linguistic influence have been found to affect both segmental and supra-segmental elements of speech.

Where the German language is concerned, despite relatively scarce research in this domain, there is some evidence that an interplay of the type of phonological errors, i.e., atypical versus typical, psycholinguistic skills, such as nonword repetition and phone imitation, and infrequent variations of

29 phones, might determine typical from atypical development in Turkish-German bilinguals (Albrecht, 2017).

1.4.3.2 Semantic deficiencies in bilingual children

The lexical difficulties of bilinguals with PLI have been substantially researched (Kohnert, 2016). Not only is there data suggesting that bilinguals have smaller receptive and expressive vocabulary in both languages compared to their typical peers as measured through picture identification and naming tasks but, also, that these limitations are evident at connected speech level (Iluz-Cohen, 2012). Furthermore, bilinguals with PLI have been shown to have a weaker ability to produce associations between

semantically linked words compared to typical peers, which is an indication of limited overall semantic knowledge. These difficulties have been found to be comparable across languages as

measured through picture description and word association tasks (Sheng et al, 2012) and there is some evidence from bilingual LI children that nouns are more easily retrieved than verbs (Kambanaros, Grohmann & Theodorou, 2013).

As mentioned in a previous chapter, bilingualism may result in increased inter- and intra-lexical errors.

It has been found that language exposure is directly linked with lexical knowledge (McMillen, Griffin, Bedore, & Oppenheim, 2020). In general, such errors, as measured through naming tasks, might be linked to increased competition between double word representations spread between the two

languages, weaker semantic links, poor linguistic skills, as well as poorer vocabularies. As regards the type of errors in each language, these are, also, dependent on the level of linguistic exposure. An interesting study revealed that children produced more semantic errors in their dominant language as opposed to more phonological ones, as well as more “I don’t know”- type of answers in their weaker

languages, weaker semantic links, poor linguistic skills, as well as poorer vocabularies. As regards the type of errors in each language, these are, also, dependent on the level of linguistic exposure. An interesting study revealed that children produced more semantic errors in their dominant language as opposed to more phonological ones, as well as more “I don’t know”- type of answers in their weaker